Multimodal data augmentation for digital twining assisted by artificial intelligence in mechanics of materials

Digital twins in the mechanics of materials usually involve multimodal data sense that an instance a mechanical component has both experimental and simulated data. These simulations aim not only to replicate observations but also extend Whether spatially, temporally, or functionally, augmentation is needed for various possible uses components improve predictions behavior. Related are scarce, high-dimensional physics-based causality relation exists between observational We propose scheme coupled with pruning, order limit memory requirements augmented This desirable digital twining assisted by artificial intelligence when performing nonlinear model reduction. Here, aims at preserving similarities terms validity domain reduced twins. In this article, we consider specimen subjected test high temperature, where as-manufactured geometry may impact lifetime component. Hence, represented twin includes 3D X-Ray tomography specimen, related finite element mesh, thermo-mechanical variables several time steps. There is, thus, each geometrical information. Multimodal data, which couple different representation modalities together, hard collect, annotating them requires significant effort. Thus, analysis generally suffers from problem scarcity. The proposed training recommending system recognizes category available set already been fully analyzed using high-fidelity models. Such enables use ROM-net fast assessment via local reduced-order